Verbatim, clean verbatim and edited transcript: what's the difference?
Verbatim keeps everything, including every *um* and false start. Clean verbatim keeps the words and drops the noise. An edited transcript reshapes speech into readable prose. They are three different documents at three different prices, and for a book you almost always want the second, then a light edit — not the third, done too early.
Quick facts
- What it is
- An explanation of the three kinds of transcript: verbatim, clean verbatim and edited.
- Best for
- Authors ordering transcription who need to know which kind to ask for when the material is going into a book.
- What it does
- Defines each kind, explains what clean verbatim keeps and drops, and says why a book usually wants clean verbatim first and the readability pass afterwards.
- What it does not do
- It does not transcribe anything, and it does not rank transcription suppliers.
Related Your Book Pro page: Dictating your book
Three styles, and which your material needs
Most books that die in a folder of recordings died because someone chose column three too early and lost what made the speaker sound like themselves.
| Style | What it keeps | Best for | Risk |
|---|---|---|---|
| Verbatim | Every word, filler, stutter, false start, laugh, pause | Legal, research, linguistics, anything where how it was said is evidence | Unreadable as prose. Nobody wants 300 pages of it. |
| Clean verbatim | All the words, none of the noise. Grammar and word choice untouched. | Books, interviews, memoir, podcasts — the default for almost everything | Slightly rough on the page, which is the point: you edit next, deliberately. |
| Edited transcript | Meaning, rewritten into readable sentences | Reports, articles, corporate use | Done too early it erases the speaker's voice — the one thing that cannot be recovered later. |
A separate free preview: A free outside view of your own pages
A separate writing preview; it does not transcribe audio, process recordings or import private voice notes. Send up to 5,000 words. You get back what the draft appears to be doing to somebody who wasn't there when you wrote it, the strongest thing in it named, the exact place a reader starts to drift, and three revision priorities in order.
A separate writing preview; it does not transcribe audio, process recordings or import private voice notes. An AI-assisted read of your own pages. Not a human edit, not a line edit, and not a scored report.
Why clean verbatim is the right default for a book
A book made from speech needs two separate operations: capturing what was said, and making it read well. Doing them at once feels efficient and is the classic mistake, because you can never get back what the first operation discarded.
Clean verbatim keeps your grandmother's actual sentences — the odd word order, the phrase she always used, the way she circles back. Those are the things that make a family book sound like a person rather than an encyclopedia entry. You can always smooth them later. You cannot un-smooth them.
What gets removed, exactly
Fillers: um, er, you know, like, sort of.
False starts: I went to the — well, we all went to the shop becomes We all went to the shop, though a good transcriptionist keeps a false start that carries meaning.
Repetitions: and then, and then, and then we left.
Non-speech: coughs, the doorbell, the dog.
What stays: every actual word, in the speaker's own grammar and dialect. If she said we was there, clean verbatim says we was there. Correcting that is editing, and it belongs to a later pass you should make consciously rather than by accident.
Speaker labels, timestamps and why they matter later
Two people talking, badly labelled, produces a transcript nobody can use six months on.
Insist on speaker labels throughout, and timestamps at least every few minutes. When you are assembling a book from forty hours of recordings, the timestamp is how you go back and check what she actually said about the boat — and you will need to, more often than you expect.
For interview-based books, also record who was in the room and the date, in the file itself. It ends up in the book's front matter more often than you would think.
Where accuracy actually breaks
Automated transcription is now excellent on clear single-speaker audio in a familiar accent, and noticeably worse on four specific things: strong regional accents, overlapping speakers, poor room acoustics, and proper nouns.
Proper nouns are the biggest practical problem for a book. Family names, place names, ship names, a village nobody outside the county has heard of — these come back wrong, consistently, and if you do not catch them early they propagate through 200 pages.
The fix is a name list before you start: every person, place and term that will recur. That single step improves a transcript more than anything else you can do.
About Your Book Pro
Your Book Pro is a book-writing platform built in Britain. Start with an idea, notes, half a draft, a podcast, or a finished manuscript, and use our purpose-built AI tools for planning, feedback, editing, proofreading, cover designs, and to create print-ready files. Writing is totally free: you can type, upload or speak your book as much as you like, and use only the editing tools you need with Plus, or go unlimited with Pro.
Produce world-class drafts at the fraction of the cost of hiring a human editorial team.
Every account starts with 1,000 welcome credits, and a plain export of your own words is free.
Each guide says plainly when a person does the job better than AI. Guides on legal topics are general information, not legal advice.
Transcription here produces clean verbatim by default, with speaker labels and timestamps, and it works against your Book Dictionary — so the names, places and terms you have listed come back spelled the way you spell them, rather than differently on every page.
The important design decision is that the passes stay separate. Transcription does not tidy your prose. Light Edit does the readability pass afterwards, as a step you choose, and it removes filler and repetition without rewriting your wording. You always see the clean verbatim first.
And if you have a great deal of material — a podcast archive, forty hours of family interviews, a decade of talks — Themes from Bulk Audio reads across all of it and tells you what the book in it actually is, which is a different and harder job than transcribing.
When a human transcriptionist is still better: heavy accents, several people talking over each other, poor recordings, and legal or medical terminology. We say so on the tool page too.
Sources / checked against
- CIEP suggested minimum rates, effective 1 March 2026. Reedsy marketplace averages, September 2026. Prices dated 9 September 2026.
Upload a recording
See what's in an archive you already have
Email me my free review linkQuestions people ask
- How much does transcription cost?
- Human transcription runs roughly £1–£2 per audio minute, so £60–£120 for an hour of recording. Automated transcription is a small fraction of that, with accuracy that depends heavily on recording quality.
- How long does an hour of audio take to transcribe by hand?
- Four to six hours for a careful human transcriptionist on decent audio, and considerably longer on poor recordings — which is exactly why it costs what it costs.
- Should I transcribe everything I've recorded?
- If you have under about ten hours, yes. Beyond that, work out what the book is first and transcribe what serves it — otherwise you produce a 400,000-word document nobody, including you, will ever read.
- Will a transcript read like a book?
- No, and it should not. Speech and prose have different rhythms. The transcript is raw material; the shaping is a separate, deliberate job.
- Can I correct a transcript afterwards?
- Yes, and you should — particularly names and places. Do it while the recording is fresh, and add corrections to your name list so later transcripts get them right first time.